Base backup missing WAL

Base backup missing WAL
Type: incident simulation · Topic: Storage & Backup · Level: L3 · Duration: 10–15 min
Launch: ride postgres start stage-05/05-base-backup-missing-wal

POSTMORTEM (root cause · how it was found · the fix · lesson)
Root cause: a base backup existed, but it was taken without WAL (-Xnone) and no
archive held the WAL it needed, so a restore could never reach a consistent
state. "The backup files are there" gave a false sense of safety — the backup was
not actually restorable.

How it was found: inspecting the backup showed its pg_wal empty — no WAL bundled,
nothing to replay to consistency.

The mitigation: take a new, self-contained base backup that streams WAL
(`pgpg action take-backup-with-wal`); its pg_wal is then populated and the backup
is restorable.

Lesson: a backup isn't valid because files exist — validate that it can reach
consistency (bundled WAL via -Xstream, or the required archived segments).
Don't ship/trust an incomplete backup; an index or CHECKPOINT is unrelated.

INVESTIGATION HINTS (the staged path to diagnose and fix)
1. A base backup exists on disk — but 'the files exist' is not 'the backup is restorable'. Inspect it (you're superuser): SELECT * FROM pg_ls_dir('/tmp/pgpg_backup/base/pg_wal'); it's empty, so the backup has no WAL to reach consistency.
2. The backup was taken without WAL (-Xnone) and there's no archive to fetch it from, so a restore could never finish recovery. The files being present is a false sense of safety.
3. Take a proper, self-contained backup: `pgpg action <session> take-backup-with-wal` (streams WAL into the backup). Then its pg_wal is populated and it's restorable. Don't trust the incomplete one, and an index/checkpoint is irrelevant.